RoseTTAFold: Accurate prediction of protein structures and interactions using a three-track neural network

Message boards : Rosetta@home Science : RoseTTAFold: Accurate prediction of protein structures and interactions using a three-track neural network

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Falconet

Send message
Joined: 9 Mar 09
Posts: 353
Credit: 1,227,479
RAC: 917
Message 102221 - Posted: 15 Jul 2021, 18:26:14 UTC
Last modified: 15 Jul 2021, 18:26:40 UTC

Paper at Science Magazine has been published and so has an article at the IPD Website.

Robetta is the public server for RoseTTAFold so it's Rosetta@home that will run jobs from researchers all over the world.

Looks like RoseTTAFold is already doing useful work - https://twitter.com/protein_artist/status/1415591642699079683
ID: 102221 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1832
Credit: 119,655,464
RAC: 11,085
Message 102226 - Posted: 16 Jul 2021, 12:02:23 UTC

It's really exciting stuff!
ID: 102226 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sean Kiely

Send message
Joined: 31 Jan 06
Posts: 65
Credit: 43,992
RAC: 0
Message 102227 - Posted: 16 Jul 2021, 19:35:13 UTC - in response to Message 102221.  

There's an excellent article at Nature that discusses both RoseTTAFold and Alphfold 2 (Deepmind). It's non-technical and very informative .

DeepMind’s AI for protein structure is coming to the masses
ID: 102227 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2125
Credit: 41,245,383
RAC: 9,571
Message 102229 - Posted: 17 Jul 2021, 11:53:59 UTC

ID: 102229 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348

Send message
Joined: 19 Jan 06
Posts: 881
Credit: 52,257,545
RAC: 0
Message 102231 - Posted: 17 Jul 2021, 13:04:29 UTC - in response to Message 102229.  

Because Baek’s and Baker’s group has released its computer code on the web, others can improve on it; the code has been downloaded 250 times since 1 July. “Many researchers will build their own structure prediction methods upon Baker’s work,” says Jinbo Xu, a computational structural biologist at the Toyota Technological Institute at Chicago. Hassabis says its computer code is now also open source. As a result of both groups’ work, progress should now be swift, Moult says: “When there’s a breakthrough like this, 2 years later, everyone is doing it as well if not better than before.”


Thanks for the cite. If you see any of these show up on the server, let me know.
It looks like they won't need us any more, though they certainly are not telling us that.
ID: 102231 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1994
Credit: 9,624,317
RAC: 7,073
Message 102232 - Posted: 17 Jul 2021, 13:50:51 UTC - in response to Message 102231.  

It looks like they won't need us any more,

I don't thins so, cause:
- these simulations needs still a lot of computational power
- these simulations is used only for one kind of simulation and not, for example, for docking


though they certainly are not telling us that.

This for sure!
ID: 102232 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348

Send message
Joined: 19 Jan 06
Posts: 881
Credit: 52,257,545
RAC: 0
Message 102233 - Posted: 17 Jul 2021, 13:59:36 UTC - in response to Message 102232.  
Last modified: 17 Jul 2021, 14:01:59 UTC

Thanks. That is more illuminating that what I have seen thus far.

When WCG/MIP shut down because they were going to AI, they gave their crunchers a month's notice.
That allowed them to migrate to other projects gracefully.

Maybe Rosetta will still have enough work, but that is not at all clear. They still have a large body of crunchers, but probably less to hand out.
ID: 102233 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1994
Credit: 9,624,317
RAC: 7,073
Message 102239 - Posted: 19 Jul 2021, 6:39:32 UTC - in response to Message 102233.  

Maybe Rosetta will still have enough work, but that is not at all clear. They still have a large body of crunchers, but probably less to hand out.


From the Ars Tecnica article:

The demand for time on these systems is likely to be intense

ID: 102239 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348

Send message
Joined: 19 Jan 06
Posts: 881
Credit: 52,257,545
RAC: 0
Message 102240 - Posted: 19 Jul 2021, 7:49:35 UTC - in response to Message 102239.  

From the Ars Tecnica article:

The demand for time on these systems is likely to be intense

That article does not mention Rosetta.
ID: 102240 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1994
Credit: 9,624,317
RAC: 7,073
Message 102242 - Posted: 19 Jul 2021, 15:01:03 UTC - in response to Message 102240.  

That article does not mention Rosetta.


But mention RoseTTAFold!! :-P
ID: 102242 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348

Send message
Joined: 19 Jan 06
Posts: 881
Credit: 52,257,545
RAC: 0
Message 102243 - Posted: 19 Jul 2021, 17:35:34 UTC - in response to Message 102242.  

That article does not mention Rosetta.


But mention RoseTTAFold!! :-P

Of course. And if they can run it faster on their own servers, they will. You would not send out work to BOINC unless you had to.
It involves parsing up the project into manageable pieces, and waiting several days to get results back before you can do another iteration if the next work unit depends on the previous ones.

Apparently some crunchers think that Rosetta was created to give them something to do. I don't think that is how Dr. Baker looks at it.
ID: 102243 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1832
Credit: 119,655,464
RAC: 11,085
Message 102244 - Posted: 19 Jul 2021, 19:03:39 UTC - in response to Message 102243.  

That article does not mention Rosetta.


But mention RoseTTAFold!! :-P

Of course. And if they can run it faster on their own servers, they will. You would not send out work to BOINC unless you had to.
It involves parsing up the project into manageable pieces, and waiting several days to get results back before you can do another iteration if the next work unit depends on the previous ones.

Apparently some crunchers think that Rosetta was created to give them something to do. I don't think that is how Dr. Baker looks at it.

The vbox application is already pumping some tasks through, and once it's better understood I expect more researchers outside the bakerlab will start using the online submission server more too.

But my guess is that the really big amounts of compute power will be needed for protein design rather than solving existing protein structures.
ID: 102244 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1832
Credit: 119,655,464
RAC: 11,085
Message 102245 - Posted: 19 Jul 2021, 19:03:41 UTC - in response to Message 102243.  
Last modified: 19 Jul 2021, 19:57:30 UTC

Duplicate post
ID: 102245 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348

Send message
Joined: 19 Jan 06
Posts: 881
Credit: 52,257,545
RAC: 0
Message 102246 - Posted: 19 Jul 2021, 20:43:59 UTC - in response to Message 102244.  

The vbox application is already pumping some tasks through, and once it's better understood I expect more researchers outside the bakerlab will start using the online submission server more too.

But my guess is that the really big amounts of compute power will be needed for protein design rather than solving existing protein structures.

I have a Ryzen 3900X with 64 GB waiting for them. Maybe they will show.
If not, I can find something else for it to do.
ID: 102246 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Falconet

Send message
Joined: 9 Mar 09
Posts: 353
Credit: 1,227,479
RAC: 917
Message 102247 - Posted: 19 Jul 2021, 20:48:32 UTC
Last modified: 19 Jul 2021, 20:58:29 UTC

I made a post in this thread and I wrote at the time that the Vbox app is TrRosetta and not RoseTTAFold - https://boinc.bakerlab.org/rosetta/forum_thread.php?id=14655&postid=102182#102182
We are already running RoseTTAFold on jobs from the Robetta server using the standard Rosetta app. I know I've had jobs on my PCs in which the "modeling method" at Robetta was RoseTTAFold - They used the Rosetta 4.20 app and Vbox wasn't used for those tasks since it's not even installed on this laptop.

I found this job at the Robetta queue- https://robetta.bakerlab.org/results.php?id=97601 - Modelling Method is RoseTTAFold.
I've also seen some jobs at Robetta that used a modeling method called "TrRefineRosetta". But I don't remember if I ran any of those - https://robetta.bakerlab.org/results.php?id=86758

All that I've seen indicates that we are running RoseTTAFold jobs using the standard Rosetta 4.20 app and we have been running those for a little while now (according to the Robetta website, RoseTTAFold has been available since the 15th of June). One of those articles on RoseTTAFold mentions that the server they set up (i.e. Robetta, IMO) had already generated more than 5000 protein structures, according to the Nature article Sean Kiely shared on this thread:

"As well as making the code for RoseTTaFold freely available, Baker’s team has set up a server into which researchers can plug a protein sequence and get a predicted structure. Since it was launched last month, the server has predicted the structure of more than 5,000 proteins submitted by around 500 people, says Baker. "

Unless I'm misunderstanding something, which is highly possible, we are indeed running RoseTTAFold using the standard Rosetta app.
ID: 102247 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jim1348

Send message
Joined: 19 Jan 06
Posts: 881
Credit: 52,257,545
RAC: 0
Message 102248 - Posted: 19 Jul 2021, 20:53:08 UTC - in response to Message 102247.  

Thanks. That was not clear to me. They all looked like ordinary 4.20 Rosetta to me.
ID: 102248 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Falconet

Send message
Joined: 9 Mar 09
Posts: 353
Credit: 1,227,479
RAC: 917
Message 102249 - Posted: 19 Jul 2021, 21:02:21 UTC - in response to Message 102248.  

One difference I noticed is the screensaver. The "stage" usually says fastrelax or fragmentmover or something like that but the RoseTTAFold jobs I saw had different stages, ones that I hadn't seen before at least.. I don't remember the names though.
ID: 102249 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1994
Credit: 9,624,317
RAC: 7,073
Message 102250 - Posted: 20 Jul 2021, 7:02:46 UTC - in response to Message 102243.  

Apparently some crunchers think that Rosetta was created to give them something to do.

Not me.
R@H was created to make science. To make science ALSO with the help of us volunteers.
At December it's 16 years for me in this project (i started with a little single core cpu), and if i can help the research in any way (cpu, gpu, virtual machines, Foldit, etc) i'm still ready!
ID: 102250 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1994
Credit: 9,624,317
RAC: 7,073
Message 102251 - Posted: 20 Jul 2021, 7:06:48 UTC - in response to Message 102247.  

Unless I'm misunderstanding something, which is highly possible, we are indeed running RoseTTAFold using the standard Rosetta app.

It would be great!
ID: 102251 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1832
Credit: 119,655,464
RAC: 11,085
Message 102252 - Posted: 20 Jul 2021, 8:34:25 UTC - in response to Message 102247.  
Last modified: 20 Jul 2021, 8:38:13 UTC

I made a post in this thread and I wrote at the time that the Vbox app is TrRosetta and not RoseTTAFold - https://boinc.bakerlab.org/rosetta/forum_thread.php?id=14655&postid=102182#102182
We are already running RoseTTAFold on jobs from the Robetta server using the standard Rosetta app. I know I've had jobs on my PCs in which the "modeling method" at Robetta was RoseTTAFold - They used the Rosetta 4.20 app and Vbox wasn't used for those tasks since it's not even installed on this laptop.

I found this job at the Robetta queue- https://robetta.bakerlab.org/results.php?id=97601 - Modelling Method is RoseTTAFold.
I've also seen some jobs at Robetta that used a modeling method called "TrRefineRosetta". But I don't remember if I ran any of those - https://robetta.bakerlab.org/results.php?id=86758

All that I've seen indicates that we are running RoseTTAFold jobs using the standard Rosetta 4.20 app and we have been running those for a little while now (according to the Robetta website, RoseTTAFold has been available since the 15th of June). One of those articles on RoseTTAFold mentions that the server they set up (i.e. Robetta, IMO) had already generated more than 5000 protein structures, according to the Nature article Sean Kiely shared on this thread:

"As well as making the code for RoseTTaFold freely available, Baker’s team has set up a server into which researchers can plug a protein sequence and get a predicted structure. Since it was launched last month, the server has predicted the structure of more than 5,000 proteins submitted by around 500 people, says Baker. "

Unless I'm misunderstanding something, which is highly possible, we are indeed running RoseTTAFold using the standard Rosetta app.


Ah that's really interesting, but confusing - the Rosetta 4.20 application is from May 2020 according to the apps page, so it's from before any Alpahfold/AI was built into Rosetta, other than TRrosetta that was a parallel project, as I understand things. Is there any evidence that these are being pushed to R@H rather than being crunched in-house?

Or maybe they are doing the first stage (the AI bit) in-house and then we are doing the final stage which uses 4.20 on R@H, which might explain the still high volume of work coming through 4.20.
ID: 102252 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · Next

Message boards : Rosetta@home Science : RoseTTAFold: Accurate prediction of protein structures and interactions using a three-track neural network



©2024 University of Washington
https://www.bakerlab.org